On 12/13/19 3:21 PM, Dean Carpenter wrote:
> On 2019-12-11 1:58 pm, Giovanni Bechis wrote:
>> On 12/11/19 3:17 PM, Bill Cole wrote:
>>> On 11 Dec 2019, at 2:39, Giovanni Bechis wrote:
>>>
>>>> On 12/11/19 6:21 AM, KADAM, SIDDHESH wrote:
>>>>> Hi PFA...
>>>>>
>>>>> On 12/11/2019 12:36 AM, Giovanni Bechis wrote:
>>>>>> On 12/10/19 7:49 PM, Michael Storz wrote:
>>>>>> [...]
>>>>>>> My copy hit
>>>>>>>
>>>>>>> BODY_SINGLE_WORD=1.347, HTML_IMAGE_ONLY_04=1.172, MPART_ALT_DIFF=0.79
>>>>>>>
>>>>>>> not enough to mark it as spammy.
>>>>>
>>>> FuzzyOcr + bayes is killing this kind of emails for me:
>>>
>>> FuzzyOcr is unmaintained and doesn't even have an authoritative repository
>>> as far as I can tell. It is computationally very expensive, to the degree
>>> that it isn't safe to just add it to an existing mail system which does not
>>> have a lot of idle CPU and memory capacity.
>>>
>> it's true that it's unmaintained but I have it running on Perl 5.28
>> with some patches and it's still useful every now and then (if you
>> have some spare cpu cycles and you know what you are doing).
>> A new ocr plugin could be definetely a better choice.
>> Giovanni
>
> fuzzyocr is available from the standard repos for Ubuntu 18.04. It's
> v3.6.0-10, with a homepage listed as
>
> https://web.archive.org/web/20130117050640/http://fuzzyocr.own-hero.net/
>
> Interestingly I just got one of those bitcoin spams, but fuzzyocr didn't pick
> up on it. This is the spam report for it :
>
If I remember well, by default fuzzyocr skips images with resolution higher
than 800x800, the spam I received had a bigger image.
Giovanni
